Ability to support file-like objects in clodius.tiles

[manzt]

Tracking whether file-like objects could be supported by the different tiles implementations.

🟩 = yes, 🟨 = unknown, 🟥 = no

🟨 bam.py

Depends on pysam.AlignmentFile which allows a file-like object for the reference but not the index.

import pysam

# internal support for remote files http://www.htslib.org/doc/samtools.html#DESCRIPTION
# We could use this, but would need to cache instances (like tiles/clodius) to avoid
# re-initializing for each call to `tiles()`
pysam.AlignmentFile('http://localhost:8080/data.bam##idx##s3://path/to/data.bai')

with open('./data.bam') as f:
  pysam.AlignmentFile(f) # works

with open('./data.bam') as f, open('./data.bai') as idx:
  pysam.AlignmentFile(f, index_filename=idx) # error, can't pass file-like object for index!

with fsspec.open('http://localhost:8080/data.bam') as f:
  pysam.AlignmentFile(f) # not recognized as a file-like object

🟥 bed2ddb.py

Depends on sqllite3.connect which only accepts a string.

🟥 bedarcsdb.py

Depends on sqllite3.connect which only accepts a string.

🟥 beddb.py

Depends on sqllite3.connect which only accepts a string.

🟩 bedfile.py

Currently a no-op.

🟨 bigbed.py

Depends on tiles/bigwig.py.

🟨 bigwig.py

Depends on whether pybbi could accept file-like objects.

🟩 chromsizes.py

Just reading a csv using builtin open.

🟩 cooler.py

Depends on h5py which allows file-like objects

🟩 density.py

Depends on h5py which allows file-like objects

🟩/🟨 fasta.py

Depends on pyfaidx.Fasta supports only filenames. Some effort here that could be abstracted further.

🟥 geo.py

Depends on sqllite3.connect which only accepts a string.

🟩 hitile.py

Depends on h5py which allows file-like objects

🟥 imtiles.py

Depends on sqllite3.connect which only accepts a string.

🟩 mrmatrix.py

Depends on h5py which allows file-like objects

🟩 multivec.py

Depends on h5py which allows file-like objects

🟩 tabix.py

Uses builtin open and gzip.open which can be replaced with f.open and gzip.open(f) if filehandles are provided.

🟩 time_interval.py

Uses builtin open with can be replaced with fh.open if file-like object is provided.

[manzt]

context: manzt/hg#6